1 / 28

Replication (14.4-14.6) by Ramya Balakumar

Replication (14.4-14.6) by Ramya Balakumar. Highly available services. A few systems that provide highly available services : Gossip Bayou Coda. The gossip architecture. the gossip architecture is a framework for implementing highly available services

eydie
Download Presentation

Replication (14.4-14.6) by Ramya Balakumar

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Replication(14.4-14.6)byRamya Balakumar

  2. Highly available services A few systems that provide highly available services : • Gossip • Bayou • Coda

  3. The gossip architecture • the gossip architecture is a framework for implementing highly available services • data is replicated close to the location of clients • RMs periodically exchange ‘gossip’ messages containing updates • gossip service provides two types of operations • queries - read only operations • updates - modify (but do not read) the state • FE sends queries and updates to any chosen RM • one that is available and gives reasonable response times • Two guarantees (even if RMs are temporarily unable to communicate • data reflects the updates seen by client, even if the use different RMs • relaxed consistency between replicas. All RMs eventually receive all updates. RMs use ordering guarantees to suit the needs of the application (generally causal ordering). Client may observe stale data.

  4. Service RM gossip RM RM Query, prev Val, new Update, prev Update id FE FE Query Val Update Clients Query and update operations in a gossip service

  5. Gossip processing of queries and updates • The five phases in performing a client request are: • request • update response - the RM replies as soon as it has seen the update • coordination • execution - the RM executes the request • query response - if the request is a query the RM now replies: • agreement • RMs update one another by exchanging gossip messages (lazily) • e.g. when several updates have been collected

  6. A gossip replica manager, showing its main state components Other replica managers Replica Replica log timestamp Gossip messages Replica manager Timestamp table Value timestamp Replica timestamp Stable Value Update log updates Executed operation table Updates OperationID Update Prev FE FE

  7. Processing of query and update operations • Query operations contain q.prev • they can be applied if q.prev ≤ valueTS (value timestamp) • failing this, the RM can wait for gossip message or initiate them • e.g. if valueTS = (2,5,5) and q.prev = (2,4,6) - RM 0 has missed an update from RM 2 • Once the query can be applied, the RM returns valueTS (new) to the FE. The FE merges new with its vector timestamp e.g. in a gossip with 3 RMs a value of (2,4,5) at RM 0 means that the value there reflects system the first 2 updates accepted from FEs at RM 0, the first 4 at RM 1 and the first 5 at RM 2.

  8. Gossip update operations • Update operations are processed in causal order • A FE sends update operation u.op, u.prev, u.id to RM i • A FE can send a request to several RMs, using same id • When RM i receives an update request, it checks whether it is new, by looking for the id in its executed ops table and its log • if it is new, the RM • increments by 1 the ith element of its replica timestamp, • assigns a unique vector timestamp ts to the update • and stores the update in its log • The RM returns ts to the FE,which merges it with its vector timestamp • For stability u.prev ≤ valueTS • That is, the valueTS reflects all updates seen by the FE. • When stable, the RM applies the operation u.op to the value,updates valueTS and adds u.id to the executed operation table.

  9. Discussion of Gossip architecture • the gossip architecture is designed to provide a highly available service • clients with access to a single RM can work when other RMs are inaccessible • but it is not suitable for data such as bank accounts • it is inappropriate for updating replicas in real time (e.g. a conference) • scalability • as the number of RMs grow, so does the number of gossip messages • for R RMs, the number of messages per request (2 for the request and the rest for gossip) = 2 + (R-1)/G • G is the number of updates per gossip message • increase G and improve number of gossip messages, but make latency worse • for applications where queries are more frequent than updates, use some read-only replicas, which are updated only by gossip messages

  10. The Bayou system • Operational transformation: Replica managers detect and resolve conflicts according to domain specific policy. • The Bayou guarantee: Eventually every replica manager receives the same set of updates; the databases of the replica managers are identical.

  11. Committed Tentative c0 c1 c2 cN t0 t1 t2 ti ti+1 Tentative update ti becomes the next committed update and is inserted after the last committed update cN. Committed and tentative updates in Bayou

  12. Bayou summary • Replication non-transparent to application • E.g: A user has booked a slot in a diary, but this slot moves to another available slot. The user comes to know of this only when he checks the diary again. • No clear indication for user as to which data is committed and which data is tentative.

  13. The Coda File system • Descendent of AFS (Andrew File System) • Vice processes—Replica Managers • Venus processes—hybrid of front ends and managers • VSG—Volume Storage Group (group of servers holding replica of a file system) • AVS—Available volume storage group • Working mostly similar to AFS

  14. Replication Strategy • Allows file modification when network disconnected or partitioned. • CVV—Coda version vector—vector timestamp with one element for each server. • Each element of CVV is an estimate of the number of modifications performed on the version of the file that is held at the corresponding server.

  15. EXAMPLE • Example: • File  F • Servers  S1, S2, S3 • Consider the modifications to F in a volume that is replicated at 3 servers. • The VSG for F is {S1, S2, S3} • F is modified at about the same time by 2 clients C1, C2. • Because of the Network fault: • C1 can access S1 and S2 only. • Therefore C1’s AVSG is {S1, S2} • C2 can access S3 only • Therefore C2’s AVSG is {S3} • Initially CVV for F is {1,1,1} at all 3 servers.

  16. Update Semantics: Guarantees offered by Coda: • After a successful open: (s’ <> NULL and (latest (F, s’, 0) or (latest(F, s’, T) and lostCallback(s’, T) and inCache(F)))) or (s’ = NULL and inCache(F)) • After a failed open operation: (s’ <> NULL and conflict(F, s’)) or (s’ = O’ and -| inCache(F)) • After a successful close: (s’ <>NULL and updated(F,s’)) or (s’ = O’) • After failed close: s’ <> NULL and conflict(F, s’)

  17. Cache Coherence: The Venus process at each client must detect the following events within T seconds of their occurrence: • Enlargement of an AVSG ( due to inaccessibility of a previously inaccessible server) • Shrinking of an AVSG (due to a server becoming inaccessible) • A lost call back event. Venus sends a probe message to all servers in the VSG of the file every T seconds.

  18. Coda File system & Bayou approach • Similar to Bayou --- Bayou used for databases and Coda for file systems • Coda conflict resolution similar to Bayou’s operational transformation approach.

  19. Client + front end Client + front end U T deposit(B,3); getBalance(A) B Replica managers Replica managers A A B B B A E.g. of Transactions on replicated data

  20. Available copies – read one/ write all available whereas Ts deposit is performed by M, N and P. T’s getBalance is performed by X At XT has read A and has locked it. Therefore U’s deposit is delayed until T finishes Client + front end T U Client + front end getBalance(B) deposit(A,3); getBalance(A) Replica managers deposit(B,3); B M Replica managers B B A A N P X Y

  21. Available copies replication: RM failure example • both RMs fail before T and U have performed their deposit operations • Therefore T’s deposit will be performed at RMs M and P (all available) • and U’s deposit will be performed at RM Y. (all available). Client + front end T U Client + front end getBalance(B) deposit(A,3); getBalance(A) Replica managers deposit(B,3); B M Replica managers B B A A N P X Y

  22. Available copies replication • Local validation (the additional concurrency control) • before a transaction commits, it checks for failures and recoveries of the RMs it has contacted • e.g. T would check if N is still unavailable and that X, M and P are still available. • If this is the case, T can commit. • this implies that X failed after T validated and before U validated • i.e. we have N fails T commits  X fails  U validates • (above, we said X fails before T’s deposit, in which case,T would have to abort) • U checks if N is still available (no) and X still unavailable • If N is unavailable :U must abort • after all the operations of a transaction have been carried out • the FE will inform the coordinator of failed RMs it knows about • the coordinator can attempt to communicate with any RMs noted to have failed • then in doing the 2PC it will discover whether any RMs involved in the transaction have subsequently failed

  23. Client + front end Client + front end Network U T withdraw(B, 4) partition deposit(B,3); B Replica managers B B B Network partition-division into subgroups

  24. Network partition Transaction T Replica managers X V Y Z Two network partitions

  25. Network partition Virtual partition Replica managers X V Y Z Virtual partition

  26. Virtual partition V Virtual partition V 1 2 Y X V Z Two overlapping virtual partitions

  27. Creating a virtual partition • Phase 1: • • The initiator sends a Join request to each potential member. The argument of Join is a proposed logical timestamp for the new virtual partition. • • When a replica manager receives a Join request, it compares the proposed logical timestamp with that of its current virtual partition. • – If the proposed logical timestamp is greater it agrees to join and replies Yes; • – If it is less, it refuses to join and replies No. • Phase 2: • • If the initiator has received sufficient Yes replies to have read and write quora, it may complete the creation of the new virtual partition by sending a Confirmation message to the sites that agreed to join. The creation timestamp and list of actual members are sent as arguments. • • Replica managers receiving the Confirmation message join the new virtual partition and record its creation timestamp and list of actual members.

  28. Summary for Gossip and replication in transactions • the Gossip architecture is designed for highly available services • it uses a lazy form of replication in which RMs update one another from time to time by means of gossip messages • it allows clients to make updates to local replicas while partitioned • RMs exchange updates with one another when reconnected • replication in transactions • primary-backup architectures can be used • other architectures allow FMs to use any RM • available copies allows RMs to fail, but cannot deal with partitions

More Related